Evaluation of the predictive and prognostic potential of blood immune cell profiles in metastatic cancer patients treated with immune checkpoint inhibitors Authors: Veera Nurmela, Anna-Mari Schroderus, Satu Tiainen, Outi Kuittinen, Olli Tenhunen, Anna Jokimäki, Sanna Suoranta, Sanna Pasonen-Seppänen, Tero Sievänen, Tuure Kinnunen, Aino Rönkä Published: 2026 ABSTRACT Immune checkpoint inhibitors (ICI) have transformed the treatment of metastatic malignancies, yet only 20–40% of unselected patients benefit. Peripheral blood could offer a minimally invasive and dynamic source for predictive biomarkers of ICI therapy.This study evaluated circulating immune cell frequencies and phenotypes before and during ICI therapy to identify tumor-agnostic blood-based biomarkers. We analyzed routine blood immune cell counts in a retrospective cohort of 202 patients treated with ICIs. Next, we investigated the findings in a prospectively collected cohort of 45 patients using multiparametric flow cytometry for characterization of immune cell subsets. We considered early radiological response, progression-free survival (PFS), and overall survival (OS) as clinical outcomes.Higher pre-treatment monocyte and lower lymphocyte counts were consistently associated with inferior PFS and OS but not with early radiological response in the retrospective cohort. In the prospective cohort, detailed immunophenotyping identified elevated pre-treatment frequencies of intermediate (CD14⁺CD16⁺) monocytes as a marker of poorer PFS and OS. Within lymphocyte subsets, higher T cell frequencies were associated with better OS, and a higher pre-treatment frequency of CD226-expressing CD8⁺ memory T cells with better ICI response. ICI therapy induced increase in TIGIT+CD8+ and CD4+ memory T cell subsets, regardless of treatment response.In conclusion, elevated pre-treatment levels of blood monocytes, together with decreased levels lymphocytes, were associated with poor survival of ICI-treated patients. Although detailed immunophenotyping of pre-treatment blood immune cell populations showed limited predictive utility, specific subpopulations, such as intermediate monocytes and CD226⁺CD8⁺ memory T cells, may harbor potential as prognostic/predictive indicators.Supplementary InformationThe online version contains supplementary material available at 10.1007/s00262-026-04404-0. Introduction Immune checkpoint inhibitors (ICIs) have emerged as the standard of care for several metastatic malignancies, with yearly expanding therapeutic indications [1]. However, only 20–40% of unselected patients benefit from the therapy, underscoring the need for predictive biomarkers [2]. Currently used biomarkers, such as PD-L1 expression or tumor mutational burden (TMB), are analyzed from tumor-tissue, and restricted access their use in clinical practice. Peripheral blood provides a dynamic and mini-invasive source of potential biomarkers, capturing signals from tumor–immune interactions while also reflecting the overall systemic immune milieu. Since the pharmacodynamic effect of ICIs is based on reviving anti-tumor immune activity, monitoring immune states from blood may serve as a surrogate for the evaluation of ICI response. Several previous studies have demonstrated that overall frequencies of major immune cell populations exert predictive value in ICI-treated patients. For example, low baseline lymphocyte levels and elevated monocyte counts have repeatedly been linked with poorer outcomes across cancer types [3, 4]. Yet the specific immune cell subsets responsible for these associations remain incompletely defined. For example, evidence regarding the widely studied CD8⁺ T lymphocyte population is still conflicting across studies [5–7], and a similar inconsistency is seen in monocyte subtypes [8, 9]. The interpretation of these findings is also complicated by considerable heterogeneity in the endpoints and biomarker measures used across studies, with some focusing on early responses and others on long-term survival, thereby obscuring whether the findings reflect predictive or merely prognostic associations. In addition to the counts of circulating immune cells, their functional characteristics may play a role in the antitumor response to ICIs. T cells are of particular interest in this regard since ICIs primarily target T-cell activation. During anti-PD-(L)1 therapy, exhausted T cells are functionally reinvigorated and thereby regain their capacity to mediate antitumor effects. Yet blockade of the PD-1/PD-L1 axis may also trigger compensatory upregulation of other coinhibitory receptors, such as T cell Ig and ITIM domain (TIGIT) or lymphocyte-activation gene 3 (LAG-3), which may contribute to ICI resistance, and provides a rationale for combined checkpoint blockade strategies [10, 11]. Longitudinal profiling of such T-cell subsets expressing co-inhibitory receptors during ICI therapy may therefore help to identify patient populations who could benefit from combination therapy approaches. In the present work, we evaluated the value of routinely measured peripheral blood immune cells as therapy biomarkers in patients treated with ICIs and examined whether detailed immune phenotyping provides additional prognostic value beyond these routine blood parameters. Material and methods Study population and materials This study included two cohorts of patients with metastatic cancers treated with ICI therapy, named retrospective and flow (cytometry) cohorts. The retrospective cohort consisted of 202 patients treated at the Cancer Center of Kuopio University Hospital, Finland, during 2015–2022. The flow cohort included 45 patients treated at the Cancer Centers of Oulu or Kuopio University Hospitals, Finland, between January 1st, 2020, and December 31st, 2023. Patients aged > 18 years with no prior ICI therapies and approved indications for ICI therapy were included. Dual ICI therapies (i.e. combinations of anti-PD(L)1 and anti CTLA4 antibodies) and ICIs combined with chemotherapy or TKIs were allowed. Age, sex, cancer type, metastatic burden and previous anti-cancer drugs used were documented as baseline demographic characteristic. The metastatic burden was assessed radiologically before ICI therapy initiation from the following sites: lymph nodes, brain, lung, liver, bone, and other. Ethics The study was conducted under the research permits from Wellbeing Services county of Northern Savo and North Osthrobonia (diary no.s 59/13.00/2023 and 27/2020) and the acceptance of the Ethical Committees of Wellbeing Services counties of these regions (diary no.s 1482/2019 and 2/2020). For the prospective flow cohort, eligible patients were identified by their treating physicians at routine oncologist appointments and recruited by a member of the research group. All participants have given written informed consent, as mandated by the Declaration of Helsinki. Peripheral blood sample collection and storage Complete blood counts, including the total leucocyte, neutrophil, monocyte and lymphocyte counts, were analyzed during routine lab blood sampling using a XN-1000 automated hematology analyzer (Sysmex). The results were retrospectively collected from electronic patient records. The blood tests were drawn 1–14 days before the first cycle of ICI treatment (pre-ICI) and 1–7 days before the second cycle of treatment (post-ICI). The blood count values are presented as x10E9/L. In the prospective cohort, blood samples for flow cytometry were collected on average four days before (pre-ICI samples) and 24 days after (post-ICI samples) the initiation of ICI therapy. Following sample collection, peripheral blood mononuclear cells (PBMCs) were isolated from heparinized peripheral blood samples using Ficoll density gradient centrifugation with SepMate tubes (StemCell) and subsequently stored in liquid nitrogen. Flow cytometry Altogether, 44 pre-ICI samples, and 41 post-ICI samples were analyzed. For analysis, PBMCs were thawed, treated with DNAse I (StemCell) according to manufacturer’s instructions, and aliquoted for surface marker staining. Six panels were used for stainings. The detailed staining protocol and panels are described in Supplemental Table 1. Representative gatings are presented in Supplemental Figs. 1, 2 and 3. Table 1Demographic characteristics of the retrospective and flow cohorts divided into therapy responders and non-responders > Retrospective cohortFlow cohortResponse statusAllResponderNon-responderAllResponderNon-respondern2028611545*1529Age (yrs) ± SD67.2 ± 10.767.9 ± 10.766.1 ± 10.767.5 ± 8.468.2 ± 8.066.9 ± 8.9GenderFemale63 (31.2)26 (30.2)37 (32.2)14 (31.1)3 (20.0)11 (37.9)Male139 (68.8)60 (69.8)78 (67.8)31 (68.9)12 (80.0)18 (62.1)Cancer typeNSCLC#96 (47.5)48 (55.8)48 (41.7)25 (55.6)12 (80.0)12 (41.4)Melanoma37 (18.3)10 (11.6)27 (23.5)12 (26.7)2 (13.3)10 (34.5)RCC##30 (14.9)11 (12.8)18 (15.7)8 (17.8)1 (6.7)7 (24.1)Other39 (19.3)17 (19.8)22 (19.1)Performance status0110 (54.5)56 (65.1)53 (46.1)20 (44.4)9 (60.0)11 (37.9) ≥ 192 (45.5)30 (34.9)62 (53.9)25 (55.6)6 (40.0)18 (62.1)Metastatic burden & sites1–2 metastatic sites140 (69.3)75 (87.2)64 (55.7)26 (57.8)12 (80.0)13 (44.8) > 2 metastatic sites62 (30.7)11 (12.8)51 (44.3)19 (42.2)3 (20.0)16 (55.2)Lymph node130 (64.4)54 (62.8)76 (66.1)31 (68.9)10 (66.7)20 (69.0)Brain20 (9.9)6 (7.0)14 (12.2)5 (11.1)2 (13.3)3 (10.3)Lung90 (44.6)41 (47.7)48 (41.7)19 (42.2)7 (46.7)12 (41.4)Liver37 (18.3)4 (4.7)33 (28.7)6 (13.3)2 (13.3)4 (13.8)Bone56 (27.7)16 (18.6)40 (34.8)10 (22.2)2 (13.3)8 (27.6)Treatment Nivolumab55 (27.2)19 (22.1)35 (30.4)13 (28.9)2 (13.3)11 (37.9)Pembrolizumab87 (43.1)34 (39.5)53 (46.1)18 (40.0)6 (40.0)11 (37.9)Atezolizumab6 (3.0)1 (1.2)5 (4.3)0 (0.0)0 (0.0)0 (0.0)Nivolumab + Ipilimumab10 (5.0)5 (5.8)5 (4.3)7 (15.6)3 (20.0)4 (13.8)ICI + chemotherapy42 (20.8)26 (30.2)16 (13.9)7 (15.6)4 (26.7)3 (10.3)Nivolumab + TKI###2 (1.0)1 (1.2)1 (0.9)0 (0.0)0 (0.0)0 (0.0)Treatment line1122 (60.4)58 (67.4)64 (55.7)35 (77.8)12 (80.0)23 (79.3) ≥ 280 (39.6)28 (32.6)51 (44.3)10 (22.2)3 (20.0)6 (20.7)Median follow-up time (months)37.6 mo41.0 mo33.8 mo28.9 mo26.0 moN/A mo*One responder was excluded from the cohort to lack of a pre-ICI blood sample#NSCLC = non-small cell lung carcinoma##renal cell carcinoma### nivolumab + tyrosine kinase inhibitorFig. 1Experimental layout and demonstration of main findings. (A). Schematic representation of the experimental layout (B). Representative gating of monocyte subtypes (C). Frequencies of CD14+CD16+ intermediate monocytes in the pre- and post-treatment samples according to the response groups (D). Overall survival of patients with high and low pre-treatment frequencies of CD14+CD16+ intermediate monocytes (E). Representative gating strategy for CD3+ T cells (F). Frequencies of T cells in pre- and post-treatment samples according to the response groups (G). Overall survival of patients with high and low pre-treatment frequencies of T cells. ICI, immune checkpoint inhibitors; PBMC, peripheral blood monocytesFig. 2Co-inhibitory and co-activation receptor expressing circulating CD8+ and CD4+ T cells in patients receiving ICI therapy. A. Heatmap of CD8+ and CD4+ T cells expressing surface receptors PD-1, TIGIT, KLRG1, CD160, CD27 and CD226 in pre- and post-treatment samples. The cell frequencies were normalized for visualization B. Pre-treatment frequencies of PD-1, TIGIT, KLRG1, CD160, CD27 and CD226 in CD8+ and CD4+ memory T cells C. Fold change of D PD-1 and E TIGIT in memory CD8+ and CD4+ T cells in responders and non-responders. Mann–Whitney U was used for statistical testing in B-E. Fold change 1 (no change) is indicated as dotted line in D-EFig. 3Heatmap of the immune cell landscape depicting immune cell frequencies pre and post ICI therapy. The cell frequencies were normalized for visualization. A Frequency of PD-1 B and TIGIT C expressing memory CD4+ and CD8+ T cells pre and post ICI treatment. Wilcoxon test was used for statistical testing in B-C. *P < 0.05, **P < 0.01, ****P < 0.0001. cDCs, classical dendritic cells; pDCs, Peripheral blood mononuclear cells; NK, natural kills Table 1Demographic characteristics of the retrospective and flow cohorts divided into therapy responders and non-responders > Retrospective cohortFlow cohortResponse statusAllResponderNon-responderAllResponderNon-respondern2028611545*1529Age (yrs) ± SD67.2 ± 10.767.9 ± 10.766.1 ± 10.767.5 ± 8.468.2 ± 8.066.9 ± 8.9GenderFemale63 (31.2)26 (30.2)37 (32.2)14 (31.1)3 (20.0)11 (37.9)Male139 (68.8)60 (69.8)78 (67.8)31 (68.9)12 (80.0)18 (62.1)Cancer typeNSCLC#96 (47.5)48 (55.8)48 (41.7)25 (55.6)12 (80.0)12 (41.4)Melanoma37 (18.3)10 (11.6)27 (23.5)12 (26.7)2 (13.3)10 (34.5)RCC##30 (14.9)11 (12.8)18 (15.7)8 (17.8)1 (6.7)7 (24.1)Other39 (19.3)17 (19.8)22 (19.1)Performance status0110 (54.5)56 (65.1)53 (46.1)20 (44.4)9 (60.0)11 (37.9) ≥ 192 (45.5)30 (34.9)62 (53.9)25 (55.6)6 (40.0)18 (62.1)Metastatic burden & sites1–2 metastatic sites140 (69.3)75 (87.2)64 (55.7)26 (57.8)12 (80.0)13 (44.8) > 2 metastatic sites62 (30.7)11 (12.8)51 (44.3)19 (42.2)3 (20.0)16 (55.2)Lymph node130 (64.4)54 (62.8)76 (66.1)31 (68.9)10 (66.7)20 (69.0)Brain20 (9.9)6 (7.0)14 (12.2)5 (11.1)2 (13.3)3 (10.3)Lung90 (44.6)41 (47.7)48 (41.7)19 (42.2)7 (46.7)12 (41.4)Liver37 (18.3)4 (4.7)33 (28.7)6 (13.3)2 (13.3)4 (13.8)Bone56 (27.7)16 (18.6)40 (34.8)10 (22.2)2 (13.3)8 (27.6)Treatment Nivolumab55 (27.2)19 (22.1)35 (30.4)13 (28.9)2 (13.3)11 (37.9)Pembrolizumab87 (43.1)34 (39.5)53 (46.1)18 (40.0)6 (40.0)11 (37.9)Atezolizumab6 (3.0)1 (1.2)5 (4.3)0 (0.0)0 (0.0)0 (0.0)Nivolumab + Ipilimumab10 (5.0)5 (5.8)5 (4.3)7 (15.6)3 (20.0)4 (13.8)ICI + chemotherapy42 (20.8)26 (30.2)16 (13.9)7 (15.6)4 (26.7)3 (10.3)Nivolumab + TKI###2 (1.0)1 (1.2)1 (0.9)0 (0.0)0 (0.0)0 (0.0)Treatment line1122 (60.4)58 (67.4)64 (55.7)35 (77.8)12 (80.0)23 (79.3) ≥ 280 (39.6)28 (32.6)51 (44.3)10 (22.2)3 (20.0)6 (20.7)Median follow-up time (months)37.6 mo41.0 mo33.8 mo28.9 mo26.0 moN/A mo*One responder was excluded from the cohort to lack of a pre-ICI blood sample#NSCLC = non-small cell lung carcinoma##renal cell carcinoma### nivolumab + tyrosine kinase inhibitor Demographic characteristics of the retrospective and flow cohorts divided into therapy responders and non-responders *One responder was excluded from the cohort to lack of a pre-ICI blood sample #NSCLC = non-small cell lung carcinoma ##renal cell carcinoma ### nivolumab + tyrosine kinase inhibitor Experimental layout and demonstration of main findings. (A). Schematic representation of the experimental layout (B). Representative gating of monocyte subtypes (C). Frequencies of CD14+CD16+ intermediate monocytes in the pre- and post-treatment samples according to the response groups (D). Overall survival of patients with high and low pre-treatment frequencies of CD14+CD16+ intermediate monocytes (E). Representative gating strategy for CD3+ T cells (F). Frequencies of T cells in pre- and post-treatment samples according to the response groups (G). Overall survival of patients with high and low pre-treatment frequencies of T cells. ICI, immune checkpoint inhibitors; PBMC, peripheral blood monocytes Co-inhibitory and co-activation receptor expressing circulating CD8+ and CD4+ T cells in patients receiving ICI therapy. A. Heatmap of CD8+ and CD4+ T cells expressing surface receptors PD-1, TIGIT, KLRG1, CD160, CD27 and CD226 in pre- and post-treatment samples. The cell frequencies were normalized for visualization B. Pre-treatment frequencies of PD-1, TIGIT, KLRG1, CD160, CD27 and CD226 in CD8+ and CD4+ memory T cells C. Fold change of D PD-1 and E TIGIT in memory CD8+ and CD4+ T cells in responders and non-responders. Mann–Whitney U was used for statistical testing in B-E. Fold change 1 (no change) is indicated as dotted line in D-E Heatmap of the immune cell landscape depicting immune cell frequencies pre and post ICI therapy. The cell frequencies were normalized for visualization. A Frequency of PD-1 B and TIGIT C expressing memory CD4+ and CD8+ T cells pre and post ICI treatment. Wilcoxon test was used for statistical testing in B-C. *P < 0.05, **P < 0.01, ****P < 0.0001. cDCs, classical dendritic cells; pDCs, Peripheral blood mononuclear cells; NK, natural kills Clinical outcome assessment The clinical outcomes assessed were radiological response, progression free survival (PFS), and overall survival (OS). The radiological response was assessed from the first contrast-enhanced computed tomography (CT) scans conducted after ICI initiation (at 8–12 weeks) by a radiology specialist using the RECIST criteria. The responses were categorized either as complete response (CR), partial response (PR), stable disease (SD), or progressive disease (PD). For data analysis, CR and PR were grouped as responders (R), while SD and PD were categorized as non-responders (NR) (Supplemental Table (Suppl. Table) 2). Progression-free survival (PFS) was defined as the time from the date of ICI initiation to either cancer progression, death from any cause or the end of follow-up (FU), whichever occurred first. Overall survival (OS) was defined from the date of ICI treatment initiation to the date of death from any cause or the end of follow-up (FU), whichever occurred first. To assess the longitudinal changes in cell frequencies and phenotypes between pre- and post-therapy time points, post-treatment cell counts divided by pre-treatment cell counts were calculated (fold change). Statistical analyses For the flow cytometry data, statistical analyses were conducted using GraphPad Prism (version 10, GraphPad Software Inc). The Wilcoxon rank sum test was used to compare pre- and post-ICI initiation samples, while Mann Whitney U test was applied for comparing responder and non-responder groups. Pairwise correlations between major blood immune cell subtypes were assessed using Spearman's rank correlation with Benjamini–Hochberg false discovery rate (FDR) correction [12]. The prospective cohort utilized for flow cytometry analyses was relatively small, and due to the exploratory nature of the analyses, correction for multiple comparisons was not performed. Survival analyses were performed in SPSS statistics software (version 27, IBM SPSS Statistics for Windows, IBM, Armonk, NY, USA) using the Kaplan–Meier method with Log Rank test. P-values < 0.05 were considered statistically significant. The patients were grouped into dichotomic PFS and OS subgroups based on immune cell cut offs determined by maximizing the log-rank test statistics in the two-group survival analysis in a Cox regression model, as previously demonstrated by Contal and O’Quigley [13]. The cut offs and the number of records and events in each subgroup are presented in the Supplemental Table 3. These cut offs were also used for NSCLC subgroup analyses (Suppl. Table 4). Cut offs based on subgroups of < 10 records were discarded. Benjamini–Hochberg false discovery rate adjusted p-values (FDR) are reported alongside uncorrected p-values, where indicated (Table 2, Suppl. Table 4, 7). The correlation heatmap was drawn with “pheatmap” R-package using in-house scripts in Rstudio [14–16]. Table 2The association of pre-treatment counts/frequencies and fold changes of blood immune cells with radiological response, PFS and OS in the retrospective and flow* study cohorts divided into A) total leukocytes, neutrophils and the subpopulations of the monocytic lineage, and B) total lymphocytes, NK, T and B cells. Statistically significant results are bolded(A) EndpointRadiological early responsePFSOSIndicatorPre-treatment median frequency/proportion (SD)Median fold change* (SD)HR (95% CI)mPFS monthsHR (95% CI)mOS monthsGrouping criterionR responderNR non-responderR responderNR non-responder > cell frequency/proportion higher than cutoff < lower than cutoff > cell frequency/proportion higher than cutoff < lower than cutoffGroupRNRpFDRRNRpFDRpFDR > ≤ pFDRpFDR > ≤ pFDRCell populationLEUCOCYTESLeukocytes(E9/L)7.05(3.86)7.00(3.62)0.780.960.99(0.26)0.96(0.39)0.560.991.03(0.99–1.07)0.180.482.2(1.8–13.2)6.5(5.3–8.5)0.020.041.07(1.02–1.12)0.0030.0214.4(10.3–20.9)24.5(19.4–31.2)0.0060.01NEUTROPHILSNeutrophils(E9/L)4.50(3.78)4.80(2.81)0.240.920.97(0.44)0.91(0.45)0.980.991.03(0.98–1.07)0.230.534.8(3.5–7.4)7.8(5.5–10.3)0.020.041.06(1.01–1.11)0.010.0413.8(9.2–19.8)26.9(20.0–31.7)0.0020.01MONOCYTIC LINEAGEMonocytes(E9/L)0.50(0.22)0.50(0.31)0.600.921.00(0.86)1.00(0.53)0.090.442.61(1.14–5.95)0.020.163.2(0.6–5.7)7.1(3.9–10.3)0.020.045.09(1.98–13.07)0.0010.027.5(6.3–8.7)22.7(17.1–28.4)0.0030.01Monocytes*22.75 (10.59)21.36 (12.52)0.980.981.08 (0.72)1.08(0.54)0.940.991.01(0.97–1.04)0.740.91N/AN/AN/AN/A1.03(1.00–1.06)0.080.2518.6(8.3–24.3)37.5(14.3–N/A)0.010.02Classical monocytes*19.6 (10.26)19.6 (11.73)0.950.981.01 (0.78)1.05(0.59)0.920.991.00(0.97–1.03)0.950.957.4(2.7–14.5)2.2 (0.8–N/A)0.160.181.03(0.99–1.06)0.120.2515.6(8.3–24.3)45.2(14.3–N/A)0.0050.01Non-classical monocytes*0.53 (0.68)0.93 (0.73)0.420.921.15 (4.49)0.97(1.23)0.110.441.55(0.96–2.51)0.070.282.6(1.6–14.3)12.2(2.8–N/A)0.040.051.33(0.79–2.24)0.280.4514.9 (5.5–37.5)24.3(20.4–N/A)0.220.23Intermediate monocytes*0.72 (0.80)1.03 (0.97)0.310.921.72 (10.10)1.13(2.66)0.070.441.43(1.01–2.04)0.040.212.7(1.9–13.2)12.4(2.8–N/A)0.030.051.60(1.10–2.31)0.010.0411.0(5.2–24.3)37.5(20.4–N/A)0.0030.01Dendritic cellscDCs*0.51 (3.27)0.47 (0.32)0.750.960.97(1.09)1.15(4.11)0.530.990.80(0.53–1.22)0.310.62N/AN/AN/AN/A0.88(0.58–1.32)0.540.72N/AN/AN/AN/ADendritic cellspDCs*0.25 (0.17)0.18 (0.14)0.580.921.10 (0.75)0.76(0.78)0.080.440.89(0.11–7.59)0.920.95N/AN/AN/AN/A0.94(0.10–8.57)0.960.96N/AN/AN/AN/A(B)EndpointRadiological responsePFSOSIndicatorPre-treatment median frequency (SD)Median fold change* (SD)HR (95% CI)mPFS monthsHR (95% CI)mOS monthsGrouping criterionR responderNR non-responderR responderNR non-responder > cell frequency/proportion higher than cutoff < lower than cutoff > cell frequency/proportion higher than cutoff < lower than cutoffGroupRNRpFDRRNRpFDRpFDR > ≤ pFDRpFDR > ≤ pFDRCell populationLYMPHO-CYTESLymphocytes(E9/L)1.70(0.70)1.60(0.76)0.510.921.00(0.24)1.00(0.24)0.800.990.82(0.64–1.04)0.090.297.4(5.5–9.6)4.5(2.9–6.5)0.0060.040.80(0.60–1.05)0.110.2524.3(16.8–31.7)18.6(13.5–23.6)0.0150.02Lymphocytes*71.5(9.73)73.9(12.87)0.520.921.04(0.23)0.96(0.21)0.250.801.0(0.97–1.04)0.840.91N/AN/AN/AN/A0.98(0.95–1.01)0.140.2537.5(7.2–N/A)19.5(13.1–26.4)0.070.09NK CELLSTotal NK cells*11.60 (7.44)12.70 (6.90)0.630.921.05 (0.86)1.09 (0.81)0.440.991.01(0.96–1.06)0.710.91N/AN/AN/AN/A0.99(0.94–1.04)0.660.75N/AN/AN/AN/ACD16 + NK cells*11.10(7.97)12.3 (6.53)0.610.921.02 (56.9)1.09 (1.32)0.410.991.01(0.96–1.06)0.600.872.8(2.0–13.2)7.7(2.5–N/A)0.370.370.99(0.94–1.04)0.710.7620.9(14.3–N/A)13.1(5.2–N/A)0.230.23CD56 + NK cells*0.14 (0.25)0.24 (0.19)0.400920.91 (0.86)0.85 (266.5)0.990.991.64(0.38–7.02)0.500.80N/AN/AN/AN/A1.96(0.45–8.51)0.370.54N/AN/AN/AN/AT CELLS*54.1 (7.48)53.1 (10.96)0.910.980.93 (0.20)0.97 (0.23)0.650.990.99(0.95–1.02)0.450.80N/AN/AN/AN/A0.97(0.94–1.01)0.130.2537.5(23.3–N/A)15.6(8.3–26.4)0.050.07B CELLS*3.68 (2.60)4.40 (4.80)0.410.920.88 (0.62)0.87 (0.50)0.760.991.12(1.02–1.23)0.020.16N/AN/AN/AN/A1.02(0.93–1.13)0.630.75N/AN/AN/AN/A*Prospective flow cytometry cohort; FDR, FDR-adjusted p-value Table 2The association of pre-treatment counts/frequencies and fold changes of blood immune cells with radiological response, PFS and OS in the retrospective and flow* study cohorts divided into A) total leukocytes, neutrophils and the subpopulations of the monocytic lineage, and B) total lymphocytes, NK, T and B cells. Statistically significant results are bolded(A) EndpointRadiological early responsePFSOSIndicatorPre-treatment median frequency/proportion (SD)Median fold change* (SD)HR (95% CI)mPFS monthsHR (95% CI)mOS monthsGrouping criterionR responderNR non-responderR responderNR non-responder > cell frequency/proportion higher than cutoff < lower than cutoff > cell frequency/proportion higher than cutoff < lower than cutoffGroupRNRpFDRRNRpFDRpFDR > ≤ pFDRpFDR > ≤ pFDRCell populationLEUCOCYTESLeukocytes(E9/L)7.05(3.86)7.00(3.62)0.780.960.99(0.26)0.96(0.39)0.560.991.03(0.99–1.07)0.180.482.2(1.8–13.2)6.5(5.3–8.5)0.020.041.07(1.02–1.12)0.0030.0214.4(10.3–20.9)24.5(19.4–31.2)0.0060.01NEUTROPHILSNeutrophils(E9/L)4.50(3.78)4.80(2.81)0.240.920.97(0.44)0.91(0.45)0.980.991.03(0.98–1.07)0.230.534.8(3.5–7.4)7.8(5.5–10.3)0.020.041.06(1.01–1.11)0.010.0413.8(9.2–19.8)26.9(20.0–31.7)0.0020.01MONOCYTIC LINEAGEMonocytes(E9/L)0.50(0.22)0.50(0.31)0.600.921.00(0.86)1.00(0.53)0.090.442.61(1.14–5.95)0.020.163.2(0.6–5.7)7.1(3.9–10.3)0.020.045.09(1.98–13.07)0.0010.027.5(6.3–8.7)22.7(17.1–28.4)0.0030.01Monocytes*22.75 (10.59)21.36 (12.52)0.980.981.08 (0.72)1.08(0.54)0.940.991.01(0.97–1.04)0.740.91N/AN/AN/AN/A1.03(1.00–1.06)0.080.2518.6(8.3–24.3)37.5(14.3–N/A)0.010.02Classical monocytes*19.6 (10.26)19.6 (11.73)0.950.981.01 (0.78)1.05(0.59)0.920.991.00(0.97–1.03)0.950.957.4(2.7–14.5)2.2 (0.8–N/A)0.160.181.03(0.99–1.06)0.120.2515.6(8.3–24.3)45.2(14.3–N/A)0.0050.01Non-classical monocytes*0.53 (0.68)0.93 (0.73)0.420.921.15 (4.49)0.97(1.23)0.110.441.55(0.96–2.51)0.070.282.6(1.6–14.3)12.2(2.8–N/A)0.040.051.33(0.79–2.24)0.280.4514.9 (5.5–37.5)24.3(20.4–N/A)0.220.23Intermediate monocytes*0.72 (0.80)1.03 (0.97)0.310.921.72 (10.10)1.13(2.66)0.070.441.43(1.01–2.04)0.040.212.7(1.9–13.2)12.4(2.8–N/A)0.030.051.60(1.10–2.31)0.010.0411.0(5.2–24.3)37.5(20.4–N/A)0.0030.01Dendritic cellscDCs*0.51 (3.27)0.47 (0.32)0.750.960.97(1.09)1.15(4.11)0.530.990.80(0.53–1.22)0.310.62N/AN/AN/AN/A0.88(0.58–1.32)0.540.72N/AN/AN/AN/ADendritic cellspDCs*0.25 (0.17)0.18 (0.14)0.580.921.10 (0.75)0.76(0.78)0.080.440.89(0.11–7.59)0.920.95N/AN/AN/AN/A0.94(0.10–8.57)0.960.96N/AN/AN/AN/A(B)EndpointRadiological responsePFSOSIndicatorPre-treatment median frequency (SD)Median fold change* (SD)HR (95% CI)mPFS monthsHR (95% CI)mOS monthsGrouping criterionR responderNR non-responderR responderNR non-responder > cell frequency/proportion higher than cutoff < lower than cutoff > cell frequency/proportion higher than cutoff < lower than cutoffGroupRNRpFDRRNRpFDRpFDR > ≤ pFDRpFDR > ≤ pFDRCell populationLYMPHO-CYTESLymphocytes(E9/L)1.70(0.70)1.60(0.76)0.510.921.00(0.24)1.00(0.24)0.800.990.82(0.64–1.04)0.090.297.4(5.5–9.6)4.5(2.9–6.5)0.0060.040.80(0.60–1.05)0.110.2524.3(16.8–31.7)18.6(13.5–23.6)0.0150.02Lymphocytes*71.5(9.73)73.9(12.87)0.520.921.04(0.23)0.96(0.21)0.250.801.0(0.97–1.04)0.840.91N/AN/AN/AN/A0.98(0.95–1.01)0.140.2537.5(7.2–N/A)19.5(13.1–26.4)0.070.09NK CELLSTotal NK cells*11.60 (7.44)12.70 (6.90)0.630.921.05 (0.86)1.09 (0.81)0.440.991.01(0.96–1.06)0.710.91N/AN/AN/AN/A0.99(0.94–1.04)0.660.75N/AN/AN/AN/ACD16 + NK cells*11.10(7.97)12.3 (6.53)0.610.921.02 (56.9)1.09 (1.32)0.410.991.01(0.96–1.06)0.600.872.8(2.0–13.2)7.7(2.5–N/A)0.370.370.99(0.94–1.04)0.710.7620.9(14.3–N/A)13.1(5.2–N/A)0.230.23CD56 + NK cells*0.14 (0.25)0.24 (0.19)0.400920.91 (0.86)0.85 (266.5)0.990.991.64(0.38–7.02)0.500.80N/AN/AN/AN/A1.96(0.45–8.51)0.370.54N/AN/AN/AN/AT CELLS*54.1 (7.48)53.1 (10.96)0.910.980.93 (0.20)0.97 (0.23)0.650.990.99(0.95–1.02)0.450.80N/AN/AN/AN/A0.97(0.94–1.01)0.130.2537.5(23.3–N/A)15.6(8.3–26.4)0.050.07B CELLS*3.68 (2.60)4.40 (4.80)0.410.920.88 (0.62)0.87 (0.50)0.760.991.12(1.02–1.23)0.020.16N/AN/AN/AN/A1.02(0.93–1.13)0.630.75N/AN/AN/AN/A*Prospective flow cytometry cohort; FDR, FDR-adjusted p-value The association of pre-treatment counts/frequencies and fold changes of blood immune cells with radiological response, PFS and OS in the retrospective and flow* study cohorts divided into A) total leukocytes, neutrophils and the subpopulations of the monocytic lineage, and B) total lymphocytes, NK, T and B cells. Statistically significant results are bolded Leukocytes (E9/L) 7.05 (3.86) 7.00 (3.62) 0.99 (0.26) 0.96 (0.39) 1.03 (0.99–1.07) 2.2 (1.8–13.2) 6.5 (5.3–8.5) 1.07 (1.02–1.12) 14.4 (10.3–20.9) 24.5 (19.4–31.2) Neutrophils (E9/L) 4.50 (3.78) 4.80 (2.81) 0.97 (0.44) 0.91 (0.45) 1.03 (0.98–1.07) 4.8 (3.5–7.4) 7.8 (5.5–10.3) 1.06 (1.01–1.11) 13.8 (9.2–19.8) 26.9 (20.0–31.7) Monocytes (E9/L) 0.50 (0.22) 0.50 (0.31) 1.00 (0.86) 1.00 (0.53) 2.61 (1.14–5.95) 3.2 (0.6–5.7) 7.1 (3.9–10.3) 5.09 (1.98–13.07) 7.5 (6.3–8.7) 22.7 (17.1–28.4) 22.75 (10.59) 21.36 (12.52) 1.08 (0.72) 1.08 (0.54) 1.01 (0.97–1.04) 1.03 (1.00–1.06) 18.6 (8.3–24.3) 37.5 (14.3–N/A) 19.6 (10.26) 19.6 (11.73) 1.01 (0.78) 1.05 (0.59) 1.00 (0.97–1.03) 7.4 (2.7–14.5) 1.03 (0.99–1.06) 15.6 (8.3–24.3) 45.2 (14.3–N/A) 0.53 (0.68) 0.93 (0.73) 1.15 (4.49) 0.97 (1.23) 1.55 (0.96–2.51) 2.6 (1.6–14.3) 12.2 (2.8–N/A) 1.33 (0.79–2.24) 24.3 (20.4–N/A) 0.72 (0.80) 1.03 (0.97) 1.72 (10.10) 1.13 (2.66) 1.43 (1.01–2.04) 2.7 (1.9–13.2) 12.4 (2.8–N/A) 1.60 (1.10–2.31) 11.0 (5.2–24.3) 37.5 (20.4–N/A) Dendritic cells cDCs* 0.51 (3.27) 0.47 (0.32) 0.97 (1.09) 1.15 (4.11) 0.80 (0.53–1.22) 0.88 (0.58–1.32) Dendritic cells pDCs* 0.25 (0.17) 0.18 (0.14) 1.10 (0.75) 0.76 (0.78) 0.89 (0.11–7.59) 0.94 (0.10–8.57) Lymphocytes (E9/L) 1.70 (0.70) 1.60 (0.76) 1.00 (0.24) 1.00 (0.24) 0.82 (0.64–1.04) 7.4 (5.5–9.6) 4.5 (2.9–6.5) 0.80 (0.60–1.05) 24.3 (16.8–31.7) 18.6 (13.5–23.6) 71.5 (9.73) 73.9 (12.87) 1.04 (0.23) 0.96 (0.21) 1.0 (0.97–1.04) 0.98 (0.95–1.01) 37.5 (7.2–N/A) 19.5 (13.1–26.4) 11.60 (7.44) 12.70 (6.90) 1.05 (0.86) 1.09 (0.81) 1.01 (0.96–1.06) 0.99 (0.94–1.04) 11.10 (7.97) 12.3 (6.53) 1.02 (56.9) 1.09 (1.32) 1.01 (0.96–1.06) 2.8 (2.0–13.2) 7.7 (2.5–N/A) 0.99 (0.94–1.04) 20.9 (14.3–N/A) 13.1 (5.2–N/A) 0.14 (0.25) 0.24 (0.19) 0.91 (0.86) 0.85 (266.5) 1.64 (0.38–7.02) 1.96 (0.45–8.51) 54.1 (7.48) 53.1 (10.96) 0.93 (0.20) 0.97 (0.23) 0.99 (0.95–1.02) 0.97 (0.94–1.01) 37.5 (23.3–N/A) 15.6 (8.3–26.4) 3.68 (2.60) 4.40 (4.80) 0.88 (0.62) 0.87 (0.50) 1.12 (1.02–1.23) 1.02 (0.93–1.13) *Prospective flow cytometry cohort; FDR, FDR-adjusted p-value Results Patient characteristics The baseline characteristics of patients in the retrospective (n = 202) and flow (n = 45) cohorts are shown in Table 1. The median ages were 67.2 years and 67.5 years, respectively. In both cohorts, there was a male dominance (68.8% and 68.9%) (Table 1). Non-small cell lung cancer (NSCLC) and melanoma were the most common types of cancer in both cohorts, followed by renal cell carcinoma (RCC). 69.3% of the patients in the retrospective and 57.8% in the flow cohort had less than three metastatic sites involved, as assessed from CT scans before ICI therapy. The metastatic burden was higher in non-responders in comparison to responders in both retrospective (> 2 metastatic sites 44.3% vs. 12.8%, p < 0.05) and flow cohorts (55.2% vs. 20.0%, p < 0.05). The PS was higher in non-responders than in responders in the retrospective cohort (PS ≥ 1 53.9% vs. 34.9%, p < 0.05). Most patients in both cohorts received single ICIs as treatment. (Table 1) In the retrospective cohort, there were more responders in the ICI + chemotherapy/TKI group (responders 31.4% vs. 14.8%) than in the patients receiving single/doublet-ICI (responders 68.6% vs. 85.2%, p < 0.05). The association of leucocyte and neutrophil counts with clinical outcomes We first analyzed the association of routine pre-treatment leukocyte subset counts with ICI therapy response and survival in the retrospective cohort. Total pre-treatment leukocyte and neutrophil counts between the responder and non-responder groups were comparable, and there were no significant alterations from pre- to post-treatment counts (fold change). However, higher pre-treatment leukocyte counts were significantly associated with poorer OS (Hazard ratio (HR) 1.07, 95% Confidence interval (CI) 1.02–1.12, p = 0.003, FDR = 0.02; mOS 14.4 vs. 24.5 mo., p = 0.006, FDR = 0.01). Similarly, higher pre-treatment neutrophil counts tended to associate with shorter PFS (HR 1.03, 95% CI 0.98–1.07, p = 0.23, FDR = 0.53; mPFS 4.8 vs. 7.8 mo., p = 0.02, FDR = 0.04) and OS (HR 1.06, 95% CI 1.01–1.11, p = 0.01, FDR = 0.04; mOS 13.8 vs. 26.9 mo., p = 0.002, FDR = 0.01) (Table 2A, Suppl. Figure 4). Of note, we conducted these analyses separately in patients with NSCLC histology (Suppl. Table 4), the largest cancer subgroup, and the findings were largely similar to the main tumor-agnostic cohort. The association of monocyte counts and their subtype frequencies with clinical outcomes In the retrospective cohort, the pre-treatment monocyte counts, and the fold changes were similar between the response groups. Higher pre-treatment monocyte counts were associated with shorter PFS (HR 2.61, 95% CI 1.14–5.95, p = 0.02, FDR = 0.16; mPFS 3.2 vs. 7.1 mo., p = 0.021, FDR = 0.04) and OS (HR 5.09, 95% CI 1.98–13.07, p = 0.001, FDR = 0.02; mOS 7.5 vs. 22.7 mo., p = 0.003, FDR = 0.01) (Table 2A). In the more detailed characterization of the monocytic cell lineage, no major differences were observed in the pre-treatment frequencies of the classical CD14+, non-classical CD16+ and intermediate CD14+CD16+ monocytes (Fig. 1C) or the DC subsets (plasmacytoid and classical DCs) between response groups, or when comparing fold changes (Table 2A). Overall, the lower frequency of all three monocyte subsets studied tended to associate with better survival (Table 2A). The strongest association was seen with intermediate (CD14+CD16+) monocytes (PFS: HR 1.43, 95% CI 1.01–2.04, p = 0.04, FDR = 0.21, mPFS 2.7 vs. 12.4 mo., p = 0.03, FDR = 0.05; OS HR 1.60, 95% CI 1.10–2.31, p = 0.01, FDR = 0.04; mOS 11.0 vs. 37.5 mo., p = 0.003, FDR = 0.01) (Table 2A, Fig. 1D). Significant associations between DC subsets and response or survival were not seen. The association of total lymphocyte counts and their subset frequencies with clinical outcomes The pre-treatment lymphocyte count did not associate with radiological response but tended to be associated with better PFS (HR 0.82, 95% CI 0.64–1.04, p = 0.09, FDR = 0.29; mPFS 7.4 vs. 4.5 mo., p = 0.006, FDR = 0.04) and OS (HR 0.80, 95% CI 0.60–1.05, p = 0.11, FDR = 0.25; mOS 24.3 vs. 18.6 mo., p = 0.002, FDR = 0.02) (Table 2B, Suppl. Figure 4). Similar tendency for OS was observed in the flow cohort (HR 0.98, 95% CI 0.95–1.01, p = 0.14, FDR = 0.25; mOS 37.5 vs. 19.5 mo., p = 0.07, FDR = 0.09). In the flow analyses, the pre-treatment frequencies of total NK, T, and B cells between the responder groups were comparable (Table 2B), as well as the frequencies of the CD16+ and CD56+ NK cell subsets, main naïve and memory T cell subsets (defined by CD45RA/CCR7/CD27 expression), or T regulatory cells (CD4+CD127lowCD25+), and main naïve and memory B cells (defined by CD27/IgD expression) (Suppl. Table 5). Small, albeit significant, alterations were observed as a slightly higher frequency of CD27−IgD− double-negative (DN) memory B cells (7.0 vs. 4.8, p = 0.0497 Suppl. Table 5) in treatment responders and a fold decrease of CD19+ unswitched memory (CD27+IgD+) B cells (1.067 vs. 0.8986, p = 0.0279); Suppl. Table 5) in non-responders. In survival analyses, a higher pre-treatment T cell frequency tended to be associated with better OS (HR 0.97, 95% CI 0.94–1.01, p = 0.13, FDR = 0.25; mOS 37.5 vs 15.6 mo., p = 0.05, FDR = 0.07) (Table 2B, Fig. 1G). Moreover, the frequency of T cells correlated negatively with the frequency of intermediate (r = − 0.53, p < 0.005) and non-classical (r = − 0.64, p < 0.0005) monocytes (Suppl. Figure 5). In Cox analysis, a higher pre-treatment B cell frequency was associated with shorter PFS (HR 1.12, 95% CI 1.02—1.23, p = 0.02, FDR = 0.16). Significant associations between NK cell populations or their subtypes with survival were not observed. The uni- and multivariate Cox regression analysis of main immune cell classes and clinical parameters Alongside laboratory parameters, age, gender, WHO PS, metastatic burden, and treatment line were selected as clinical parameters for Cox regression analyses (Table 3). In univariate analyses, significant associations between leucocytes, neutrophils and monocytes and survival were seen, as previously presented in Table 2A. PS was associated with poorer PFS (HR 1.80, 95% CI 1.31–2.46, p < 0.001) and OS (HR 2.28, 95% CI 1.60–3.25, p < 0.001), along with metastatic burden (PFS HR 2.61, 95% CI 1.88–3.63, p < 0.001; OS HR 2.38, 95% CI 1.67–3.39, p < 0.001, Table 3A). Significant associations in the univariate analysis were chosen for multivariate analysis. In multivariate analysis (Table 3B), the clinical variables WHO PS (PFS: HR 2.25, 95% CI 1.39–3.65, p < 0.001; OS: HR 1.97, 95% CI 1.17–3.30, p = 0.010) and metastatic burden (PFS: HR 2.20, 95% CI 1.35–3.58, p = 0.002; OS: HR 1.77, 95% CI 1.07–2.94, p = 0.027) were significantly associated with survival. Out of blood cell counts, only monocytes retained significance for OS (HR 4.56, 95% CI 1.07–19.51, p = 0.041). Table 3(A) Uni- and (B) multivariate Cox regression analysis in the retrospective cohort including clinical and laboratory pre-treatment variables. Statistically significant results are boldedA)UnivariatePFSOSHR (95% Cl)pHR (95% Cl)pGender (0 female, 1 male)0.82 (0.58–1.15)0.2440.77 (0.53–1.13)0.178Age0.99 (0.98–1.00)0.2520.99 (0.97–1.00)0.097PS (WHO 0 vs. ≥ 1)1.80 (1.31–2.46) < 0.0012.28 (1.60–3.25) < 0.001Metastatic burden (1–2 vs. ≥ 3)2.61 (1.88–3.63) < 0.0012.38 (1.67–3.39) < 0.001Treatment line(0 vs. ≥ 1)1.28 (0.94–1.75)0.1211.34 (0.95–1.89)0.096Leukocytes (E9/L)1.03 (0.99–1.07)0.1821.07 (1.02–1.12)0.003Neutrophils (E9/L)1.03 (0.98–1.07)0.2251.06 (1.01–1.11)0.011Monocytes (E9/L)2.61 (1.14–5.96)0.0235.09 (1.98–13.07) < 0.001Lymphocytes (E9/L)0.82 (0.64–1.04)0.0990.80 (0.60–1.05)0.110B) MultivariatePFSOSHR (95% Cl)pHR (95% Cl)pPS (WHO 0 vs. ≥ 1)2.25 (1.39–3.65) < 0.0011.97 (1.17–3.30)0.010Metastatic burden (1–2 vs. ≥ 3)2.20 (1.35–3.58)0.0021.77 (1.07–2.94)0.027Leukocytes (E9/L)0.92 (0.68–1.24)0.5780.58 (0.58–1.10)0.174Neutrophils (E9/L)1.06 (0.81–1.39)0.6941.29 (0.97–1.73)0.082Monocytes (E9/L)3.23 (0.85–1.24)0.0844.56 (1.07–19.51)0.041 Table 3(A) Uni- and (B) multivariate Cox regression analysis in the retrospective cohort including clinical and laboratory pre-treatment variables. Statistically significant results are boldedA)UnivariatePFSOSHR (95% Cl)pHR (95% Cl)pGender (0 female, 1 male)0.82 (0.58–1.15)0.2440.77 (0.53–1.13)0.178Age0.99 (0.98–1.00)0.2520.99 (0.97–1.00)0.097PS (WHO 0 vs. ≥ 1)1.80 (1.31–2.46) < 0.0012.28 (1.60–3.25) < 0.001Metastatic burden (1–2 vs. ≥ 3)2.61 (1.88–3.63) < 0.0012.38 (1.67–3.39) < 0.001Treatment line(0 vs. ≥ 1)1.28 (0.94–1.75)0.1211.34 (0.95–1.89)0.096Leukocytes (E9/L)1.03 (0.99–1.07)0.1821.07 (1.02–1.12)0.003Neutrophils (E9/L)1.03 (0.98–1.07)0.2251.06 (1.01–1.11)0.011Monocytes (E9/L)2.61 (1.14–5.96)0.0235.09 (1.98–13.07) < 0.001Lymphocytes (E9/L)0.82 (0.64–1.04)0.0990.80 (0.60–1.05)0.110B) MultivariatePFSOSHR (95% Cl)pHR (95% Cl)pPS (WHO 0 vs. ≥ 1)2.25 (1.39–3.65) < 0.0011.97 (1.17–3.30)0.010Metastatic burden (1–2 vs. ≥ 3)2.20 (1.35–3.58)0.0021.77 (1.07–2.94)0.027Leukocytes (E9/L)0.92 (0.68–1.24)0.5780.58 (0.58–1.10)0.174Neutrophils (E9/L)1.06 (0.81–1.39)0.6941.29 (0.97–1.73)0.082Monocytes (E9/L)3.23 (0.85–1.24)0.0844.56 (1.07–19.51)0.041 (A) Uni- and (B) multivariate Cox regression analysis in the retrospective cohort including clinical and laboratory pre-treatment variables. Statistically significant results are bolded Treatment line (0 vs. ≥ 1) The association of the expression of co-inhibitory and co-activation receptors by T cells with clinical outcomes As anti-PD(L)1 immunotherapy may lead to the upregulation of other co-inhibitory receptors as a mechanism of ICI resistance [17, 18], we next used flow cytometry to analyze the expression of co-inhibitory/co-activation receptors PD-1, TIGIT, KLRG1 (Killer cell lectin-like receptor G1), CD160, CD27, CD226, Tim-3 and LAG-3 on CD8+ and CD4+ memory T cells at baseline in treatment responders and non-responders (Fig. 2A). Tim-3 was minimally expressed by circulating T cells, as expected [19], as well as LAG-3 and CD160 by CD4+ memory T cells, and therefore these markers were not analyzed further (Suppl. Table 6). A higher frequency of CD8+ memory T cells expressing the CD226 co-activation marker was observed in treatment responders (89.6% vs. 83.0%, p = 0.04; Fig. 2B). Otherwise, comparable pre-treatment frequencies and fold changes of co-inhibitory receptor-expressing CD8+ and CD4+ memory T cells were seen between the responder groups (Fig. 2A-E). In survival analyses, a higher pre-treatment frequency of cells expressing the co-activation molecule CD226+ tended to be associated with better survival in both CD4+ (mPFS 16.6 vs. 2.7 mo., p = 0.05, FDR = 0.13; mOS 26.8 mo. vs. 13.6 mo., p = 0.06, FDR = 0.14) and CD8+ memory T cell compartments (mPFS 12.8 vs. 8.7 mo., p = 0.11, FDR = 0.14, mOS 26.8 vs. 13.1 mo., p = 0.02, FDR = 0.14) (Suppl. Table 7, Suppl. Figure 6). The overall effect of ICI therapy on circulating immune cells Finally, to examine the overall pharmacological effect of ICI treatment on the immune cell landscape, we studied major immune cell frequencies regardless of the response status between pre- and post-treatment samples. No major differences were observed in the monocyte, DC or lymphocyte frequencies between pre- and post-treatment samples indicating that the immune cell subsets did not significantly deviate from baseline at the post-ICI timepoint (Fig. 3A, Suppl. Table 6). Only small, albeit significant, alterations within the lymphocyte compartment were observed, with a lower frequency of naïve CD8+ (p = 0.002) and total CD4+ T cells (p = 0.03), and a higher frequency of CD27− TEMRA CD8+ T cells (p = 0.005) and CD27+ EM CD4+ T cells (p = 0.02) in the post-treatment compared to pre-treatment samples, regardless of the response groups (Suppl. Table 6). When we studied the expression of PD-1, TIGIT, KLRG1, CD160, CD27, and CD226 in CD8+ and CD4+ memory T cells between the pre- and post-treatment samples, PD-1 expression was, as expected, significantly reduced in response to anti PD-(L)1 therapy in both CD8+ and CD4+ T cells (both p < 0.0001) (Fig. 3B, Suppl. Table 6). Small but significant increases in TIGIT+ CD8+ (p = 0.015) and TIGIT+ CD4+ memory T cells (p = 0.002) were also observed between the pre- and post-treatment samples (Fig. 3C, Suppl. Table 6). Discussion In this study, we assessed the potential of routinely measured major blood immune cell counts and their subtypes to predict the response and survival of patients with advanced cancer undergoing ICI therapy. Higher total pre-treatment monocyte counts, and lower lymphocyte counts were associated with shorter survival. Using more detailed immune cell subtyping, decreased survival was associated particularly with higher intermediate monocyte and lower T cells frequencies. Notably, strong associations between major immune cell populations, or their extensively studied subsets including monocyte, NK, and B/T lymphocyte subtypes, and early radiological ICI response remained unobserved. Thus, rather than predicting early responders to ICIs, blood immune cell characterization may serve as a tool for stratifying long term survivors after ICI therapy. A relative increase in myeloid cells with a concomitant decrease in lymphoid cells in peripheral blood has been consistently associated with poorer survival of patients with cancer in general [20–22], and reduced benefit from ICI therapy across tumor types [3, 4]. Our findings from the retrospective cohort were concordant with these studies and were further confirmed in an independent prospective flow cohort, in which we also observed a strong inverse correlation between monocyte and lymphocyte counts. Biologically, the observations most likely reflect the systemic inflammation related to the overall burden of the malignant disease [23] and are therefore predominantly prognostic in nature rather than predictive of ICI therapy. Further supporting this interpretation, the clinical variables reflecting disease burden retained independent prognostic significance in our multivariable Cox regression analysis, while the associations of blood counts with survival were largely attenuated. Our flow subtyping of monocytes suggested that poorer prognosis was specifically associated with higher pretreatment frequencies of intermediate (CD14+CD16+) monocytes. Intermediate monocytes exhibit immunosuppressive properties functionally distinct from classical or non-classical monocytes [24] and appear to harbor M2 pro-tumorigenic phenotype with the capacity to infiltrate tumor-areas in patients with cancer [25]. Our current findings are consistent with a previous study by de Lima et al. [9] and support the concept that increased intermediate monocytes are linked with impaired T cell function which may attenuate ICI efficacy [9, 26]. However, results contrasting our study have also been reported [8, 27], and some studies have failed to identify any associations between monocytes and survival outcomes [6]. The discrepancies may be explained by the plasticity of monocytes, resulting in functional heterogeneity already at baseline and further phenotypic shifts during treatment [28]. Additionally, differences in cohort composition, heterogenicity in tumor histologies, methodological approaches, and sampling schedules are likely to have contributed to the inconsistent findings. In the future, larger prospective studies with well-defined cohorts will be needed to clarify the heterogeneity observed. The characterization of lymphocyte subsets in our study revealed a trend towards longer OS in patients with elevated baseline T cell frequencies. These results are in line with prior studies where lower lymphocyte counts have been associated with progression [29] and a robust T-cell pool with prolonged survival [30]. We did not see clear associations between pre-treatment B cell and NK cell subset frequencies and OS nor signals of early increase in lymphocyte counts during ICI therapy—findings that contrast some of the previous studies [5, 31]. We may have missed very early, possibly transient immune cell dynamics due to a later collection of post-treatment samples (approx. 24 days after ICI initiation), and we also lacked cell proliferation markers from our flow staining panel. Since resistance to ICI can be conveyed by not only PD-1 blockade but also other coinhibitory receptors, we wanted to investigate how PD-1 blockade affects the expression of other co-stimulatory/inhibitory receptors on circulating memory CD8+ and CD4+ T cells. When comparing pooled pre- and post-treatment samples, we found a decrease in the frequency of PD-1 expressing CD8+ and CD4+ memory T cells after therapy initiation, consistent with effective PD-1 receptor blockade in vivo. We also found an increased frequency of CD4+ and CD8+ memory T cells expressing TIGIT in the post-treatment samples. However, since these expression patterns were similar between responders and non-responders, they most likely represent general pharmacodynamic effects of PD-1 blockade rather than associate with biological efficacy of ICIs. The expression of other co-inhibitory receptors examined showed also minimal discriminatory capacity between responder groups. We observed, however, a slightly higher baseline frequency of CD8+ memory T cells expressing the co-activation marker CD226 in ICI responders, and trending signals towards longer survival in patients with higher baseline frequencies of both CD4+ and CD8+ memory T cells expressing CD226. Downregulation of CD226 on CD8+ T cells in tumors has previously been described as a potential mechanism of immune escape [34, 35]. Conversely, the increased expression of CD226 on CD8+ T cells appears to induce tumor destruction and predict response during PD-1 blockade [36], indicating the potential of CD226 expression as a predictive biomarker for ICI therapy. Our current results are in line with these papers but need further validation in a larger cohort. Overall, our extensive immune profiling did not identify strong cell populations discriminating early radiological responders from non-responders. This may partly stem from the limitations of RECIST-based response criteria in immunotherapy, where radiological response does not always translate into meaningful survival benefit. This has been stated also in previous studies [32, 33]. ICI-associated phenomena such as pseudo-progression and delayed treatment responses can confound early radiological assessments of therapeutic efficacy, particularly when compared with the more rapidly manifesting effects of conventional chemotherapy. Indeed, in our study, patients exhibiting early radiological responses were more frequently treated with chemo-immunotherapy combinations rather than ICI monotherapy. Furthermore, upon subsequent CT imaging, 13% (6/45) of early non-responders in the flow cohort converted to responders (data not shown). Collectively, these findings suggest that early radiological responses may disproportionately reflect the chemotherapy-driven effects, while delayed immunotherapy responses may be misclassified as disease progression at early time points. The strengths of the current study include the tumor-agnostic approach, combined use of a large retrospective cohort, and a separate, prospectively collected cohort that was more comprehensively immunophenotyped. Sample processing and flow cytometry analyses were conducted using standardized protocols while minimizing technical variability. The limits of this study include the relatively small flow cohort size, heterogeneity in tumor types and therapy regimens, which all limit the statistical power and complicate the interpretation of immune-biomarker associations. Importantly, while tumor heterogeneity increases generalizability, it also introduces significant heterogeneity that may confound biomarker associations. Additionally, the identified associations in the flow cohort are preliminary and require validation in larger, independent cohorts, specifically powered for these endpoints. Moreover, noise from systemic conditions unrelated to cancer may have failed to capture potentially relevant immune cell signals in the analyses. In conclusion, our findings indicate that elevated pre-treatment levels of blood monocytes combined with reduced lymphocyte levels are tumor-agnostically associated with inferior survival outcomes in patients treated with ICIs. Specific subpopulations, such as intermediate monocytes and CD226⁺CD8⁺ memory T cells, may harbor potential as prognostic/predictive indicators and warrant further investigation. While detailed immunophenotyping of blood immune cell populations yields important biological insights into immune cell dynamics during ICI therapy, it provided limited incremental prognostic value beyond readily available major immune cell frequencies derived from routine blood counts. Supplementary Information Below is the link to the electronic supplementary material.Supplementary file1 (PDF 2548 KB) Supplementary file1 (PDF 2548 KB)